#02 Quality
We support smart quality control and management for high-end metal production, taking advantage of IIoT and Big Data technologies. With dimensional and quality in-line inspection and measurement systems, as well as software solutions for real-time monitoring and control, customers can slash costs and optimise yield, through early detection of potential non-conformities and prompt implementation of corrective actions.
Improve plant quality and monitor it in real time.
Case
History
#01
Who
A plant in the steel industry producing special steel in the form of long products to meet the engineering requirements of the automotive, earth-moving and agricultural machinery, energy, mechanics and construction sectors.
Why
The customer needed to:
- decrease the number of non-conformities and customer complaints;
- track and manage process deviations through the production chain;
- reduce the learning curve for the newly-installed SBQ rolling mill;
- identify the root cause of quality issues, prepare and follow up corrective and preventive action plans;
- help develop an analytical mindset in quality and production staff.
How
Installation and setup of Q3-PREMIUM, a Decision Intelligence suite for data-driven quality control and management based on an Industrial IoT platform dedicated to the metals industry.
The ingestion module collects and integrates process and quality data flows from various sources throughout the process chain, enabling traceability of the entire product lifecycle and genealogy.All information processed by the system is made available to the quality engineer through an Enterprise Web Portal with specific features for online quality control and configurable dashboards for in-depth statistical data analytics.
Machine learning models can be built in a collaborative MLOps environment and deployed in production for real-time quality prediction and decision support.On top of this platform, predictive use cases are being developed to address specific issues, such as forecasting mechanical properties and root cause analysis for surface defects on cast blooms and rolled bars.
Process knowledge is combined with insights generated by data science projects, forming a custom rules management system. The objective of this is to provide an online statistical process control system and automated product grading at each quality checkpoint in the production flow.
The system guides the user through the process, from early detection of potential issues to the generation of proactive decision support strategies for the implementation of remedial and preventive measures to manage quality non-conformities.
What
The customer benefited thanks to:
- increased control of all process steps;
- reduced influence of human-induced variability;
- predictive alerting of the defects-occurrence risk;
- real-time automatic grading of product quality;
- support for compliance management and quality certification integrated with the MES solution.
- improved traceability of product history from a single source;
- availability of all quality-related data in great detail;
- formalisation of internal know-how in sets of rules;
- consolidation of process setup practices thanks to data-driven insights.
Who
A plant in the steel industry producing special steel in the form of long products to meet the engineering requirements of the automotive, earth-moving and agricultural machinery, energy, mechanics and construction sectors.
Why
The customer needed to:
- decrease the number of non-conformities and customer complaints;
- track and manage process deviations through the production chain;
- reduce the learning curve for the newly-installed SBQ rolling mill;
- identify the root cause of quality issues, prepare and follow up corrective and preventive action plans;
- help develop an analytical mindset in quality and production staff.
How
Installation and setup of Q3-PREMIUM, a Decision Intelligence suite for data-driven quality control and management based on an Industrial IoT platform dedicated to the metals industry.
The ingestion module collects and integrates process and quality data flows from various sources throughout the process chain, enabling traceability of the entire product lifecycle and genealogy.All information processed by the system is made available to the quality engineer through an Enterprise Web Portal with specific features for online quality control and configurable dashboards for in-depth statistical data analytics.
Machine learning models can be built in a collaborative MLOps environment and deployed in production for real-time quality prediction and decision support.On top of this platform, predictive use cases are being developed to address specific issues, such as forecasting mechanical properties and root cause analysis for surface defects on cast blooms and rolled bars.
Process knowledge is combined with insights generated by data science projects, forming a custom rules management system. The objective of this is to provide an online statistical process control system and automated product grading at each quality checkpoint in the production flow.
The system guides the user through the process, from early detection of potential issues to the generation of proactive decision support strategies for the implementation of remedial and preventive measures to manage quality non-conformities.
What
The customer benefited thanks to:
- increased control of all process steps;
- reduced influence of human-induced variability;
- predictive alerting of the defects-occurrence risk;
- real-time automatic grading of product quality;
- support for compliance management and quality certification integrated with the MES solution.
- improved traceability of product history from a single source;
- availability of all quality-related data in great detail;
- formalisation of internal know-how in sets of rules;
- consolidation of process setup practices thanks to data-driven insights.
Case
History
#02
Who
A plant producing stainless steel in the form of flat products for a wide range of applications such as automotive, manufacturing, consumer goods, catering, construction and design. The plant covers the entire manufacturing cycle, from meltshop and finishing lines to hot rolling and cold processing.
Why
The main goal was to reduce the amount of slabs subject to grinding after casting and to optimise process practices in order to decrease the level of non-conformance resulting from defects on semi-products.
How
The specific use case implemented during the study concerns prediction of surface defects on stainless steel coils, done by analysing casting process variables acquired during slab production.
First of all, the cast slabs were logically linked to their respective child coils. Next, all available process variables associated with those slabs during casting were used as features. The defects identified on the coils were used as labels and different binary classification models were used, by means of a supervised learning approach. The machine learning models were tested in two modes: by using data aggregated on the entire slab and by splitting the slab into detailed slices half a metre long.
The use case was developed by the customer in concert with the University of Perugia, a centre of excellence in data science. The so-called “Team Data Science Process” was used. This is an agile, iterative data science methodology for efficiently delivering predictive data analytics solutions while fostering team spirit and learning.
What
The customer benefited thanks to:
- optimisation of overall product grading as a result of its new capacity to obtain valuable information on final quality ahead of time;
accumulation of additional knowledge for identification of the metallurgical causes of slivers defects.
Who
A plant producing stainless steel in the form of flat products for a wide range of applications such as automotive, manufacturing, consumer goods, catering, construction and design. The plant covers the entire manufacturing cycle, from meltshop and finishing lines to hot rolling and cold processing.
Why
The main goal was to reduce the amount of slabs subject to grinding after casting and to optimise process practices in order to decrease the level of non-conformance resulting from defects on semi-products.
How
The specific use case implemented during the study concerns prediction of surface defects on stainless steel coils, done by analysing casting process variables acquired during slab production.
First of all, the cast slabs were logically linked to their respective child coils. Next, all available process variables associated with those slabs during casting were used as features. The defects identified on the coils were used as labels and different binary classification models were used, by means of a supervised learning approach. The machine learning models were tested in two modes: by using data aggregated on the entire slab and by splitting the slab into detailed slices half a metre long.
The use case was developed by the customer in concert with the University of Perugia, a centre of excellence in data science. The so-called “Team Data Science Process” was used. This is an agile, iterative data science methodology for efficiently delivering predictive data analytics solutions while fostering team spirit and learning.
What
The customer benefited thanks to:
- optimisation of overall product grading as a result of its new capacity to obtain valuable information on final quality ahead of time;
accumulation of additional knowledge for identification of the metallurgical causes of slivers defects.
